#Releasing a project

1 messages · Page 1 of 1 (latest)

wide chasm
#

I think this is very difficult because we have 2 targets and it's not the same performing a release for Erlang than for JavaScript, even if we want to get a project which is generating a part for JavaScript in the front-end and a part for the backend, and we want the backend is in Erlang I guess it's easier to get two different projects instead of a mono-repo, but if we focus on these two options,

  • how we perform the release for a JavaScript project which could be downloaded for the browser to be run?
  • how we perform the release for a BEAM system which could start an application, supervision tree and act as a backend?
quick gull
#

What do you mean by releasing? Publishing to hex?

frail bolt
#

@wide chasm i think erlang 'Releases' are not directly translated in gleam..

#

however, that doesn't mean you cand to a build and run it as software.

wide chasm
wide chasm
# frail bolt however, that doesn't mean you cand to a build and run it as software.

I saw in the documentation that is intended to run that inside of a docker container, but I usually use my server and I prefer don't put everything inside of containers, I don't know why but that's decreasing the performance, I had to put all of the BEAM projects I have out of Docker and they are running directly using a systemd configuration

#

I guess that using gleam run it could be a solution and the configuration could be included as environment variables, it's not taking advantage of the OTP subsystem but it could work

rich zinc
#

maybe gleam export erlang-shipment can help you?

quick gull
#

I use export erlang-shipment

#

We'll get releases one day

wide chasm
#

neat! thanks!

frail bolt
#

Manual , in the release tests that my company runs. Cold start on metal compared to cold start in container there is less than 10,0000 clock cycles different in container. I attribute that to cache behaviour at this point.

#

I would not consider the kernels handling of namespaces to be a factor in container performance

#

I have caused bigger regression across the rhel platform by leaving in a printk, fortunately it never made it to customers.

#

I would like to link or share benchmarks, but I would need to talk to the redhat legal team, and they would spend 12 weeks trying to determine if sharing it would give a competitor some advantage.

wide chasm
#

I used containers last time mainly for compilation and I stopped doing that because it was the difference between 2 minutes (bare-metal), 10 minutes (VM), 90 minutes (containers) 😐 ... I think it's because Docker is configuring a very bad filesystem and it's not a good choice for compilation, but all the times I tried to use Docker or containers, it's showing me it's not good for hard data manipulation in disk, but I'm not saying that it's not working for having a Nginx, or any other stateless server, while it's not touching disk 😄

frail bolt
#

that might be more related to the supported filesystem ( i think thats aufs for docker). I want to say that it was overlay2 for rhel.

#

however i'd need to dig up info to find that.

#

which i'm sure has its on share of complexities.